A Humanoid Robot for Hand-Sign Recognition in Human-Robot Interaction (HRI)

摘要

Using a set of features, a 3-D representation of the angular relationship between a dozen biological components is created. Hidden Markov models are then applied to a feature vector in order to encode it. An in-depth explanation of the process of developing a transition gesture model is presented in order to accurately recognize significant movements. Model reduction, which combines similar states using data-dependent statistics and relative entropy to lower the number of states in the transition gesture model, is used to reduce the number of states in the model. A posture or stance is defined by the static hand combinations and hand postures that are used. In this paper, we investigate hand gesture detection using data from time-of-flight sensors and existing machine learning methods. The findings demonstrate that hand postures may be recognized with a minimum of inaccuracy when using a computer. According to the experimental results, the system successfully recognizes hand motions, and the system's performance is suitable for real-time implementations.